UC Berkeley — KLPS-4 Survey

Published: August 2017

Note: this page summarizes the rationale behind a GiveWell Incubation Grant to the Regents of the University of California, Berkeley. Staff associated with the KLPS-4 survey reviewed this page prior to publication.

Summary

As part of GiveWell's work to support research to further our understanding of our current priority programs, the Regents of the University of California, Berkeley (“UC Berkeley”) received a GiveWell Incubation Grant of $1,104,259 to support the Kenya Life Panel Survey Round 4 (KLPS-4), a follow-up study to assess the long-term impact of deworming on consumption. This survey will follow up on the study that is most central to GiveWell's current view of deworming. We hope that the results will provide a useful update to the evidence base for deworming, which could influence the amount of money that we recommend donating to deworming charities.

Table of Contents

Background

The studies that play the most significant role in supporting GiveWell's current recommendation of deworming charities are a series of long-term follow-ups to a study (Miguel and Kremer 2004) that find that deworming led to large income gains for a subset of the population. For more information, see this blog post and our deworming intervention report. These studies suggested that wages for males increased about 10 years after deworming (which is the main finding underlying our recommendation). The main way in which we hope that the KLPS-4 study will add value is by measuring consumption (rather than only wages) in a large sample (~5,000-6,000 individuals) about 20 years after deworming.

Independently of this gift, further follow-up studies are planned on whether wages increased for men 20 years after deworming, so we would have access to those results without providing any additional funding. However, we believe that the consumption data measured through our gift to support KLPS-4 could provide additional valuable information. Wages may not accurately estimate the financial wellbeing of all individuals in the sample, since many jobs in the developing world are informal. For example, someone who runs a farm might have very little formal income that would be captured through existing measurements of “wages,” but his or her financial wellbeing could be measured through a consumption survey.

About the gift

The KLPS-4 project is being led by Professor Ted Miguel and Dr. Joan Hamory Hicks, who plan to use the gift to survey roughly 5,000-6,000 people 20 years after the initial Miguel and Kremer 2004 study to determine whether deworming had an effect on their consumption. The survey will include two "waves" of roughly 2,500-3,000 people. Professor Miguel expects to have results from Wave 1 by mid-2018 and results from Wave 2 by mid-2019. The study is powered to detect roughly a 14% gain in consumption per capita.1

The gift will be used to fund the data collection activity (personnel and transport to conduct survey interviews) and personnel to analyze the data.

We recommended this gift because it supports important research that is relevant to one of GiveWell's top priority programs. The results from the study could substantially influence the amount of money that we recommend donating to deworming charities. At the end of 2016, four of our seven top charities focused on deworming.

Plans for follow-up

We plan to follow up with the gift recipient roughly every six months to check in on the timeline for receiving results from this study. At this stage, our understanding is that Wave 1 results will be available by mid-2018 and Wave 2 results will be available by mid-2019. We are uncertain when results will be able to be shared publicly, but aim to write publicly about the results as soon as we are able to.

We also plan to follow up with the recipient to share their pre-analysis plan publicly and, when the study is completed, to share data publicly.

Internal forecasts

We’re experimenting with recording explicit numerical forecasts of events related to our decisionmaking (especially grantmaking). The idea behind this is to pull out the implicit predictions that are playing a role in our decisions, and make it possible for us to look back on how well-calibrated and accurate those are. For this gift, we are recording the following forecasts:

  • 25% chance that the KLPS-4 survey significantly positively updates us on deworming, i.e. finds a result that increases our estimated cost-effectiveness for deworming by at least 2x.
  • 5% chance that the KLPS-4 survey significantly negatively updates us on deworming, i.e. finds a result that decreases our estimated cost-effectiveness for deworming by at least 2x.
  • 1

    Professor Miguel told us the following about the power of the study:

    "We assume the same distribution of consumption per capita as in KLPS-3 (the data we shared with your group), and the same intra-cluster correlation (0.037). In the full proposed sample of 5-6k, the MDE (with standard 80% power and 5% significance level) is 14 log points, i.e., roughly a 14% gain in consumption per capita."